Hard drives
From ThorxWiki
(Difference between revisions)
(slight expand) |
(.) |
||
| (6 intermediate revisions by one user not shown) | |||
| Line 1: | Line 1: | ||
| − | testing on harddrives.... |
+ | {{TOCright}} |
| − | * Bring strong [[Magnets]] externally to drive and test for bitrot. Same without case? |
+ | == Measuring drive performance == |
| + | * A basic Linux utility to measure drive latency and throughput, and record these continuously. |
||
| + | ** http://www.coker.com.au/bonnie++/readme.html |
||
| + | ** http://linuxappfinder.com/utilities/benchmarks |
||
| + | |||
| + | |||
| + | == Thrash the drive == |
||
| + | Drive thrashing should involve the looping creation (write) and md5sum (read) and then deletion of files of random sizes. Data from /dev/zero |
||
| + | * Is this efficient really? Do we need random filesizes? How about a range of constant sizes? Or just constant sizes? |
||
| + | |||
| + | == Destructive testing == |
||
| + | Test a drive and then stress it till the tests break. Keep going till the drive breaks... |
||
| + | |||
| + | # setup computer with an OS and boot drive. |
||
| + | # attach drive-to-test |
||
| + | # format drive |
||
| + | ## consider two partitions, then read/write files on both to observe :D |
||
| + | # write data, md5sum written data (/dev/zero, or urandom data? Random size or predictable size? predictable sized zero data can have pre-known md5sums) |
||
| + | # apply test |
||
| + | # md5sum data again to check against bit corruption |
||
| + | # consider 'spamsum' to check for how MUCH of the data is corrupted. |
||
| + | |||
| + | * consider applying all this in a loop which runs constant, generating files (/proc/partitions), md5sum, test, md5sum again, delete and repeat |
||
| + | * 10meg files? |
||
| + | |||
| + | === [[Magnets]] === |
||
| + | I have alot of nice strong rare-earth magnets (from hard drives...) |
||
| + | * test for effects of magnetic field when writing files as well as reading |
||
| + | * test the effect of a moving magnetic field, vs a static one (as applied both before and after a format of the drive?) |
||
| + | |||
| + | === foreign materials === |
||
* open breather hole, apply smoke |
* open breather hole, apply smoke |
||
| − | * open breather hole, but apply coke alternate: drill hole) |
+ | ** thrash drive - how long till observable issues? |
| − | * open drive entirely simply to watch movement :) |
+ | * open breather hole, but apply liquid (water or coke) |
| − | ** apply soot to drive platters |
+ | ** will it seize up the drive? Or froth up the coke? |
| + | * Drill hole - shavings do what now? |
||
| + | * sparkler (inverted, through pre-drilled hole) |
||
| + | * spray paint (after opening drive)** chalk spray |
||
| + | ** metallic spray |
||
| + | ** apply soot |
||
| + | |||
| + | |||
| + | == let's get physical... == |
||
| + | * you put your hard drive in, you pull your hard drive out, you put your hard drive in and you shake it all about |
||
| + | ** (not innuendo) |
||
| + | * melt solder to short or remove components |
||
| + | * hit drive with a hammer (rubber, then metal) |
||
| + | * drop drive (in as far as is possible when attached to cables) |
||
| + | * open drive entirely... :) |
||
Latest revision as of 15:47, 15 December 2014
|
[edit] Measuring drive performance
- A basic Linux utility to measure drive latency and throughput, and record these continuously.
[edit] Thrash the drive
Drive thrashing should involve the looping creation (write) and md5sum (read) and then deletion of files of random sizes. Data from /dev/zero
- Is this efficient really? Do we need random filesizes? How about a range of constant sizes? Or just constant sizes?
[edit] Destructive testing
Test a drive and then stress it till the tests break. Keep going till the drive breaks...
- setup computer with an OS and boot drive.
- attach drive-to-test
- format drive
- consider two partitions, then read/write files on both to observe :D
- write data, md5sum written data (/dev/zero, or urandom data? Random size or predictable size? predictable sized zero data can have pre-known md5sums)
- apply test
- md5sum data again to check against bit corruption
- consider 'spamsum' to check for how MUCH of the data is corrupted.
- consider applying all this in a loop which runs constant, generating files (/proc/partitions), md5sum, test, md5sum again, delete and repeat
- 10meg files?
[edit] Magnets
I have alot of nice strong rare-earth magnets (from hard drives...)
- test for effects of magnetic field when writing files as well as reading
- test the effect of a moving magnetic field, vs a static one (as applied both before and after a format of the drive?)
[edit] foreign materials
- open breather hole, apply smoke
- thrash drive - how long till observable issues?
- open breather hole, but apply liquid (water or coke)
- will it seize up the drive? Or froth up the coke?
- Drill hole - shavings do what now?
- sparkler (inverted, through pre-drilled hole)
- spray paint (after opening drive)** chalk spray
- metallic spray
- apply soot
[edit] let's get physical...
- you put your hard drive in, you pull your hard drive out, you put your hard drive in and you shake it all about
- (not innuendo)
- melt solder to short or remove components
- hit drive with a hammer (rubber, then metal)
- drop drive (in as far as is possible when attached to cables)
- open drive entirely... :)
